scikit.md

Scikit Example

Here's a simple example project where we used wandb with sklearn. Try running the example notebook and see results on the open project page.

import numpy as np
from sklearn import datasets
from sklearn.preprocessing import StandardScaler
from sklearn.model_selection import train_test_split
from sklearn.svm import SVC
import wandb


# Initialize wandb
# In this example we're sending runs to an open project I set up at
# app.wandb.ai/wandb/iris
wandb.init(entity="wandb", project="iris")

# Set and save hyperparameters         
wandb.config.gamma = 0.5
wandb.config.C = 1.8
wandb.config.test_size = 0.3
wandb.config.seed = 0

iris = datasets.load_iris()

X = iris.data[:, [2, 3]]
y = iris.target

X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=wandb.config.test_size, random_state=wandb.config.seed)

sc = StandardScaler()
sc.fit(X_train)

X_train_std = sc.transform(X_train)
X_test_std = sc.transform(X_test)

X_combined_std = np.vstack((X_train_std, X_test_std))
y_combined = np.hstack((y_train, y_test))

# Fit model
svm = SVC(kernel='rbf', random_state=wandb.config.seed, gamma=wandb.config.gamma, C=wandb.config.C)
svm.fit(X_train_std, y_train)

# Save metrics
wandb.log({"Train Accuracy": svm.score(X_train_std, y_train), 
           "Test Accuracy": svm.score(X_test_std, y_test)})

# Create a matplotlib custom plot to save 
def plot_data():
    from matplotlib.colors import ListedColormap
    import matplotlib.pyplot as plt

    markers = ('s', 'x', 'o')
    colors = ('red', 'blue', 'lightgreen')
    cmap = ListedColormap(colors[:len(np.unique(y_test))])
    for idx, cl in enumerate(np.unique(y)):
        plt.scatter(x=X[y == cl, 0], y=X[y == cl, 1],
               c=cmap(idx), marker=markers[idx], label=cl)

    wandb.log({"Data": plt})

plot_data()